> i agree about wealth being the primary driver of energy use
That's not what I meant. I was referring to the development of populations. Populations that didn't have what the global north would call 'the basics', electricity, household water/sanitation, lighting, adequate heating/cooling, refrigeration, as well as 'basic luxuries' like television/computing/internet and personal transport, increasingly now have access to these technologies (even in 'underdeveloped' countries). This is going to increase energy (not just electricity) usage. But once 'the basics' have been met, they will already be in line with the efficiency standards adopted by the west (maybe even moreso because energy is more expensive in the global south)
So, a counterargument might be that AI will become part of our 'basic' lifestyle, and that we will see a resurgence of demand again, but we're seeing that cloud based compute is acceptable for the vast majority. So even though, say, internet search was energy intensive at its inception, it has largely been amortized and itself energy optimized to not really raise the bar of energy usage. Custom silicon, tighter semiconductor nodes, newer algorithms, photonics, and eventually yet-to-be-discovered technologies like room temperature super conductors can again bring the energy usage down for compute, increasing efficiency not just for AI, but across the board.
that's what i'm talking about too, but yeah, i don't agree that there's a finite set of 'basics'; i think it's dependent on how much you can afford. for 50 years that's changed only very slowly in rich countries, and now it's about to change faster than it did 200 years ago, not because of the kinds of slow efficiency improvement at the point of energy consumption, but because of photovoltaic panels
things that already exist but could get much cheaper due to cheaper energy might include
ai, as you suggest, but also personal computers, oocyte cryopreservation, ecm machining, space travel, weekly air travel, personal helicopters, atmospheric carbon capture, caribbean cruises, making things out of aluminum or titanium instead of plastic, cnc machining rather than casting or stamping, cars, buildings, photovoltaic panels themselves, etc. and presumably there are other things that haven't been invented yet because they'd be uneconomic
That's not what I meant. I was referring to the development of populations. Populations that didn't have what the global north would call 'the basics', electricity, household water/sanitation, lighting, adequate heating/cooling, refrigeration, as well as 'basic luxuries' like television/computing/internet and personal transport, increasingly now have access to these technologies (even in 'underdeveloped' countries). This is going to increase energy (not just electricity) usage. But once 'the basics' have been met, they will already be in line with the efficiency standards adopted by the west (maybe even moreso because energy is more expensive in the global south)
So, a counterargument might be that AI will become part of our 'basic' lifestyle, and that we will see a resurgence of demand again, but we're seeing that cloud based compute is acceptable for the vast majority. So even though, say, internet search was energy intensive at its inception, it has largely been amortized and itself energy optimized to not really raise the bar of energy usage. Custom silicon, tighter semiconductor nodes, newer algorithms, photonics, and eventually yet-to-be-discovered technologies like room temperature super conductors can again bring the energy usage down for compute, increasing efficiency not just for AI, but across the board.